Per-tenant AI cost cap is on by default
Every tenant now ships with a $2 hard cost cap per agent run, configurable upward in ai-tool-policies. The loop terminates when the cap is hit. No surprises in the Stripe bill.
Changelog · Monthly product updates
Every change Grid makes to the platform, dated and explained. Drop your email below to get the monthly summary.
Get the monthly changelog by email.
One short email a month. No tracking pixels. Unsubscribe by replying.
Every tenant now ships with a $2 hard cost cap per agent run, configurable upward in ai-tool-policies. The loop terminates when the cap is hit. No surprises in the Stripe bill.
You can now scrub a single agent run through three nested logs — agent run, tool invocation, model call. Filter by user, tenant, or rule. Export to CSV in one click.
An agent can now spawn a sub-agent that can spawn its own sub-agent. Cost and step caps inherit. The parent gets a flattened trace at the end.
{{contact.firstName}} and the rest of the email-template set now resolve on outbound WhatsApp messages. Unresolved variables stay literal so you catch them before send.
Recurring tasks scheduled in a DST-observing timezone could drift by an hour after the change. Tasks now anchor to the wall-clock time you set, not the absolute UTC instant.
Agencies running multiple client tenants can now bill each client through a separate Stripe Customer. Parent admin still manages everything from one console. White-label per child.
The iOS app now syncs incrementally using hybrid logical clocks. Offline edits reconcile cleanly when you reconnect. Daily briefing delivered at your hour.
Workflows can now run on task / contact / deal create, update, or delete. Scheduled runs and wait conditions are first-class. Named events let one workflow chain into another.
Snoozed threads now return to Inbox at the time you set, with the hasUnread flag flipped. Five-minute polling cadence — fast enough to catch a same-day snooze.
New write tools for crmDealsUpdate, crmActivitiesCreate, and tasksDraftFromTranscript. Every tool is tenant-scoped via the F900 invariant and logged through the three-lens audit.
Auditor selection in progress. Projected completion Q4 2026. The Security page tracks current posture per framework — we move items from "in progress" to "ready" only after the auditor signs.
Every tenant can request a full export — CRM, tasks, files, audit log — in standard formats. If you ever leave Grid, your data leaves with you.
Sending a quote from a CRM record sometimes detached the resulting email from the thread it came from. Replies now land back on the same record automatically.
No tracking pixels. No newsletter spam. One email a month.